- Email: [email protected]
Monitoring Breast Cancer Response to Neoadjuvant Systemic Chemotherapy Using Parametric Contrast-Enhanced MRI: A Pilot Study
Monitoring Breast Cancer Response to Neoadjuvant Systemic Chemotherapy Using Parametric Contrast-Enhanced MRI: A Pilot Study1 Chen-Pin Chou, MD, Ming-Ting Wu, MD, Hong-Tai Chang, MD, Yu-Shin Lo, MD, Huay-Ben Pan, MD Hadassa Degani, PhD, Edna Furman-Haran, PhD
Rationale and Objectives. Neoadjuvant systemic therapy (NST) is the standard treatment for locally advanced breast cancer and a common option for primary operable disease. It is important to develop standardized imaging techniques that can monitor and quantify response to NST enabling treatment tailored to each individual patient, and facilitating surgical planning. Here we present a high spatial resolution, parametric method based on dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI), which evaluates breast cancer response to NST. Materials and Methods. DCE-MRI examinations were performed twice on 17 breast cancer patients, before and after treatment. Seven sets of axial breast images were sequentially recorded at 1.5 Tesla applying a three-dimensional, gradient echo at a spatial resolution ⬃2 ⫻ 1.2 ⫻ 0.6 mm3 and temporal resolution ⬃2 minutes, using gadopentate dimeglumine (0.1 mmol/kg wt). Image analysis was based on a color-coded scheme related to physiologic perfusion parameters. Results. A high Pearson correlation coefficient of 0.96 (P ⬍ .0001) was found between the histopathologic estimation of viable neoplastic tissue volume and the segmented volume of all the pixels demonstrating fast and steady state washout after NST (colored in light red and green). Segmentation of these pixels before and after NST indicated response in terms of reduced tumor volume and a parallel decrease in enhancement rate which reflects diminished transcapillary transfer of the contrast agent. Conclusions. The use of a parametric MRI technique provided a means to standardize segmentation and quantify changes in the perfusion of breast neoplastic tissue in response to NST. Whether this technique can serve to predict breast cancer recurrence and survival rates requires further clinical testing. Key Words. Breast cancer; MRI; neoadjuvant systemic therapy; prognosis. ©
AUR, 2007
Acad Radiol 2007; 14:561–573 1
From the Departments of Radiology (C.-P.C., M.-T.W., H.-B.P.), Surgery (H.T.C.), and Pathology and Laboratory Medicine (Y.-S.L.), Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ROC; School of Medicine, National Yang Ming University, Taipei, Taiwan, ROC (C.-P.C., M.-T.W., H.-T.C., Y.-S.L., H.-B.P.); Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel (H.D., E.F.-H.). Received January 27, 2007; accepted February 7, 2007. Supported by a research grant from Kaohsiung Veterans General Hospital, VGHKS-91– 84, an Israel Science Foundation grant (number 801/04) and by Lord David Alliance, United Kingdom. Address correspondence to: E.F.H. e-mail: edna.haran@ weizmann.ac.il
© AUR, 2007 doi:10.1016/j.acra.2007.02.005
Neoadjuvant systemic therapy (NST) is the standard treatment for locally advanced breast cancer and a common option for primary operable disease. It is mainly designed to reduce tumor size, thereby improving surgical outcomes; to evaluate response to systemic therapy; and to obtain long-term, disease-free survival (1). To achieve these goals, it is necessary to develop standardized imaging techniques that can noninvasively monitor the response of breast tumors to NST and quantify changes in their size and spread, as well as track specific biologic and physiologic markers of malignancy. Such imaging techniques could be useful in the early stages of treatment, to help predict response to chemotherapy, to enable
561
CHOU ET AL
treatment tailored to each individual patient, and to facilitate surgical planning at later treatment stages. Magnetic resonance imaging (MRI), an important adjuvant tool for the detection and characterization of breast cancer (2– 4) has also been used to monitor the effects of NST (5). Specifically, contrast-enhanced MRI provides a means to delineate the architectural and dynamic features of breast tumors and determine their size (6). In recent years, several groups have examined whether contrastenhanced MRI is accurate enough to evaluate residual tumor size after NST, and compared their results with final histopathologic assessment of mastectomy specimens (7–11) and with data from other imaging methods such as mammography and sonography. In addition to detecting variations in tumor size, dynamic contrast-enhanced (DCE) MRI can provide information regarding the pathophysiologic response of the tumor vasculature to NST. It is well-known that tumor angiogenesis leads to the formation of blood microvessels that are excessively permeable and enhance leakage of bloodborne contrast agents, leading to augmented contrast enhancement. In tumors responding to chemotherapy, tumor angiogenesis is halted, and the development of necrosis and fibrosis leads to the establishment of a microcapillary network with properties different from that feeding the growing tumor. These changes can be quantified by analyzing the enhancement parameters of dynamic contrast-enhanced images (12–15) or by using tracer pharmacokinetic models that calculate transcapillary transfer constants and other related parameters (8,16 –18). Modelbased DCE analysis both before and after the first or second chemotherapeutic treatment has also proven useful in predicting its final outcome, and may aid in designing the type and timing of therapy (19 –22). Thus, in addition to changes in tumor size, variations in the transcapillary transfer constants can help to identify patients likely to be nonresponsive to a particular therapy, early in treatment. Although some studies have suggested that MRI evaluation before breast surgery can decrease local recurrence rates after breast conservation (23), the potential to clarify the clinical picture for patients undergoing NST, still has some controversial issues (24,25). Because tumors are usually heterogeneous in nature, it is important to evaluate their response to neoadjuvant systemic therapy at high spatial resolution. However, in practical terms, it is difficult to achieve images with sufficient signal-to-noise at the temporal resolution dictated by the necessity for high spatial resolution to optimize the precision of the parameters derived from the fitting procedures
562
Academic Radiology, Vol 14, No 5, May 2007
required for pharmacokinetic modeling. In this article, we present a pilot study demonstrating the use of a high-resolution, parametric method based on DCE MRI to evaluate response to NST. This method was originally developed to improve breast cancer diagnosis (26,27) and, in that context, demonstrated high sensitivity and specificity in differentiating between benign and malignant breast lesions (28). It was also employed in a mouse model of human breast cancer xenografts to monitor response to hormonal therapy (29). In the present study, we demonstrate that this technique enables the design of an objective segmentation procedure for quantifying changes in tumor volume and the spread of the disease in the breast, as well as estimating variations in microvascular function within viable neoplastic regions.
MATERIALS AND METHODS Patients, Treatment, and Timing of MRI A total of 17 patients, 35– 62 years of age (median age, 48 years) were recruited to this study, which was carried out between 2001 and 2003. All of the patients signed an informed consent form approved by the institutional review board of the hospital. Patients were initially diagnosed with Stage II-III infiltrating ductal carcinoma, using incision biopsy or core needle biopsy, and were scheduled for NST. Other requirements for entry into the study included patient age (⬍70), lack of pregnancy, and demonstration of adequate hematologic, renal, hepatic, and cardiac function. Patients received two to four courses of treatment with 5-fluorouracil (500 mg/m2), epi-Adriamycin (70 mg/m2), and cyclophosphamide (500 mg/m2), at 3-week intervals. Breast MRI scans were performed twice, before NST commenced and after it concluded, before surgery. In one case, a technical failure precluded analysis of the images recorded before the start of chemotherapy. Histologic Evaluation After NST, patients underwent a mastectomy and then, the breast specimens (n ⫽ 16) were sectioned at 5-mm increments, perpendicular to the cranial-caudal axis, and stained with hematoxylin and eosin for histopathologic evaluation. This evaluation was conducted by the same experienced pathologist with no prior knowledge of preoperative MRI findings. In one case, the breast specimen was not evaluated systematically in this way and hence was not included in the correlation analysis with the MRI
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 1. Color-coded calibration map of the “three time point” (3TP) parametric images. (a) A color-coding scheme of the various enhancement patterns using selected imaging sets at three time points. The washout pattern is coded by color hue: red signifies a decline in signal intensity from the first to the second postcontrast time points; green, no change in signal intensity in the first and second postcontrast time points, within an average noise level; and blue, an increase in signal intensity from the first to the second postcontrast time points. The washin rate is denoted by color intensity (in arbitrary units ranging from 0 to 256) and normalized to the maximum intensity in the calibration map. (b) A calibration map calculated using the experimental conditions of this study. (c) Simulated enhancement curves of the white dots in (b) for (ktrans,ve) ⫽ (0.9, 0.6) in red and (ktrans,ve) ⫽ (0.2, 0.4) in blue. The broken black lines in (c) demonstrate the deviation of the estimated initial rates determined by the 3TP software, from the initial rates simulated on the basis of the Tofts model (31).
findings. Pathologic evaluations of various tissue volumes in the residual tumor bed, and, within each volume, the fraction of viable neoplastic cells, fibrosis, and necrosis, were performed using a grid morphometric analysis. The fractions of the various tissue types within the residual mass also included scattered microscopic cancerous foci. MRI All of the examinations were performed using a 1.5Tesla MRI scanner equipped with a two-channel, phased array breast coil (GE Medical Systems, Waukesha, WI). MRI of both breasts were acquired in the axial plane by using a three-dimensional gradient echo sequence without fat suppression with echo time/repetition time ⫽ 4.2/7.9 milliseconds; flip angle ⫽ 30°; field of view ⫽ 32– 40 cm; matrix ⫽ 64 ⫻ 256 ⫻ 512 interpolated to 120 ⫻ 512 ⫻ 512, with a final slice thickness of 1.0 mm. Seven image sets, each taken over approximately 2 minutes, were sequentially recorded. The contrast agent, gadopentate dimeglumine (Magnevist, Schering, Berlin, Germany), was injected intravenously in the middle of the second image set by means of an automated injection system (Optistar MR injector, Mallinckrodt, Hazelwood, MO), using a dose of 0.1 mmol/kg body weight at a flow rate of 3 mL/second, followed by a 10-mL saline flush. Image Analysis Images were transferred to a Linux-operated personal computer and processed at pixel resolution using in-house
developed algorithms, as well as the “three time point” (3TP) pharmacokinetic software (26). This software is based on the Tofts physiologic model (30,31) that relates the time evolution of contrast enhancement in a tissue, to the influx transcapillary transfer constant, kin, and the efflux transcapillary transfer constant, kep. In turn, kep is equal to the ratio of the outward transcapillary transfer constant, and the extravascular, extracellular volume fraction, kout/ve. The 3TP software selects three sets of images, the first recorded before injection of the contrast agent (corresponding to time point zero) and the second and third recorded postcontrast (corresponding to the first and second time points, postcontrast). Each pixel was then coded by means of color hue and color intensity (Fig 1). Color hue reflects changes in signal intensity between the second and third time points, and is related to the washout pattern of the contrast agent: fast clearance is coded red; steady-state clearance coded green; and continuous entrance of the contrast agent with no clearance, coded blue. Each color hue is presented in arbitrary units ranging from 0 to 256, and normalized to the maximum intensity in a calibration map (see next paragraph), reflecting the rate of change in signal intensity between the first and second time points (apparent initial rate) (Fig 1). The execution time of the 3TP algorithm on a standard P4 2.8 GHz PC was approximately 30 seconds for each case. To achieve standardized utilization of this color-coded scheme, it was necessary to select two specific postcontrast time points. This selection is based on construction
563
CHOU ET AL
of a calibration map that simulates the Tofts model (30,31), and takes into account the specific set of the imaging parameters recorded (type of sequence, echo time, repetition time, flip angle), T1 and T2 relaxivities, the plasma pharmacokinetic parameters of the contrast agent, and the T1 relaxation rate of the imaged tissue, which depends on the magnetic field strength of the scanner. In building the calibration map, we assumed that the influx and outflux transcapillary transfer constants are equal; hence, the y-axis corresponds to ktrans (kin ⫽ kout ⫽ ktrans) and the x-axis to ve (Fig 1b). For any two selected postcontrast time points and for each pair of ktrans and ve values, the software algorithm calculates theoretical signal intensities, and assigns a color hue and color intensity according to the scheme depicted in Fig 1a, yielding a color-coded calibration map standardized for the sequence parameters and contrast agent (Fig 1b). Thus the combination of color hue and color intensity in a parametric image reflects a specific combination of ktrans and ve values in the calibration map (Fig 1b), which determines the changes in concentration of the contrast agent over time (Fig 1c). Based on previous estimations of ktrans and ve values in infiltrating breast cancer (32) and on the contrast agent and field strength employed in our study, the following three time points were chosen for all parametric analyses: precontrast (time point zero), 2 minutes postcontrast, and 6 minutes postcontrast. To exclude normal glandular enhancement from the 3TP parametric analysis of the neoplastic tissue, we initially selected all the pixels that exhibited more than 30% enhancement in the first postcontrast scan. Using this threshold selection criterion and normalizing the maximal intensity by means of the calibration map yielded an arbitrary color intensity scale ranging from 104 to 256. We then segmented the parametric images by drawing a region of interest around the whole breast in every fifth consecutive slice (at intervals of 5 mm) similar to the grid morphometric analysis. In each case, we quantified the number of pixels exhibiting each color hue, both before and after chemotherapy. This enabled us to assess changes in the number of pixels of each color hue and hence changes in the volume of each pixel group induced by chemotherapy. The assessment of viable neoplasm before and after NST was based on the change in volume of red ⫹ green pixels on the 3TP parametric images. The response to NST was assessed in each patient according to the published Response Evaluation Criteria in Solid Tumors (33). We also generated distribution histograms
564
Academic Radiology, Vol 14, No 5, May 2007
of color intensity for each pixel group, and calculated the median value of this parameter, both before and after chemotherapy. Statistical Evaluation Correlation between pathology data and the analyses of the 3TP parametric images according to color hue was assessed using the Pearson correlation test and two-tailed, paired Student’s t-test. A Bland-Altman plot was employed to further evaluate whether the parameters obtained from analysis of the parametric images either underestimated or overestimated the corresponding histologically determined parameters. The paired Student’s t-test was also applied to evaluate chemotherapy-induced changes in color hue and color intensity of segmented pixels in the 3TP parametric images. All statistical analyses were carried out using SAS software (version 8.0 for Windows).
RESULTS Our objectives in this pilot study were twofold: to extend the 3TP parametric method, based on high-resolution, contrast-enhanced MRI to monitor the response of malignant breast tumors to neoadjuvant systemic therapy and to quantify, in a standardized manner, the changes it induced. Accordingly, 17 patients diagnosed with infiltrating ductal breast cancer and scheduled to receive NST were scanned twice, before and after treatment, and before mastectomy. Detailed histopathologic evaluation of the entire breast in axial planes, similar to those scanned by MRI, was conducted after mastectomy in 16 patients (1 patient underwent surgery and final histologic evaluation at another hospital). Of these 16 cases, 15 were diagnosed with infiltrating ductal carcinoma, and 1 with a mixture of infiltrating ductal carcinoma, and ductal carcinoma in situ. None of the tumors showed complete pathologic response to the treatment. For each patient, the histopathologic features of mastectomy specimens and prognostic factors and the response to NST and disease recurrence over time are summarized in Table 1. Grid morphometric analysis yielded an estimate of the volume of the entire lesion postchemotherapy, as well as the percentage from this volume of viable neoplastic regions, fibrosis, and necrosis (Table 1). Before NST, the postcontrast images of the diseased breasts revealed marked signal enhancement, compared to
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Academic Radiology, Vol 14, No 5, May 2007
Table 1 Data from Mastectomy Specimens, Lymph Node Status, and Follow up % Viable NST Final Residual Recurrence-free Neoplastic Case No. Morphologic Pathologic Lymph Node Survival, Cells† No. Stage Cycle Distribution Volume† (cm3) % Fibrosis† % Necrosis† Response Involved Months 1 2 3 4 5 6 7 8 9 10 11 12 13 14* 15 16 17
III III II III III III III II II III II III III II III III II
3 3 4 3 3 3 3 3 2 3 3 4 2 2 3 3 3
Localized Localized Localized Diffused Localized Localized Localized Localized Diffused Diffused Localized Diffused Localized Diffused Diffused Diffused Diffused
NA 1.2 4.6 5.5 43.3 5.3 27.9 6.5 2.8 14.7 2.5 47.6 4.4 46.1 60.9 3.2 155
NA 4 59 22 15 47 70 62 47 76 25 62 78 5 47 94 10
NA 95 25 75 80 50 25 34 50 20 62 33 20 75 20 5 50
NA 1 16 3 5 3 5 4 3 4 13 5 2 20 33 1 40
PR PR PR PR PR PR PR SD SD SD SD SD SD SD SD SD SD
0 5 0 0 6 0 3 0 0 5 0 22 0 0 0 2 1
45 22‡ 33 44 39 50 9‡ 50 54 38 44 6‡ 44 38 12‡ 53 48
NST: neoadjuvant systemic therapy; NA: not available; PR: partial response; SD: stable disease (defined using Response Evaluation Criteria in Solid Tumors). *Mixed diagnosis: infiltrating ductal carcinoma and ductal carcinoma in situ. †Derived from grid morphometric analysis of residual tumor mass. ‡Recurrence.
the precontrast images (Fig 2a, first row, and Fig 3a, first row). After NST, enhancement patterns differed among patients, showing either localized (Fig 2c, first row) or scattered changes (Fig 3c, first row). Application of the 3TP software yielded color-coded parametric images (Fig 2a,c, second row; Fig 3a,c, second row) enabling the automated segmentation of pixels over a region of interest of the entire breast, according to their color hue and intensity. In all patients, the 3TP parametric images obtained before NST were characterized by a large fraction of red and green pixels, reflecting a fast or steady-state washout from high efflux transcapillary transfer rates and a small fraction of blue pixels, reflecting delayed washout from low efflux transcapillary transfer rates (Fig 2, 3). After NST, the total number of enhanced pixels, and hence colored pixels (red ⫹ green ⫹ blue), changed in each patient. The large volume fraction of red and green pixels in all lesions, as well as previous studies indicating the utility of a washout pattern for identifying tumor tissue (13,14) suggested that these pixels could be used as markers for segmenting viable tumor regions. We tested this hypothesis by correlating the color segmentation results of the red and green pixels (Table 2) with the histo-
logic evaluations after NST (Table 1). This analysis indicated a high Pearson correlation coefficient of 0.96 (P ⬍ .0001) between the combined volumes of red ⫹ green pixels after NST and the pathologic estimation of cellular neoplastic volume, as shown in Fig 4a. Furthermore, a Bland-Altman plot (Fig 4b) indicated no bias in the direction of either over- or underestimation of the volume of viable tumor tissue, when the color segmentation method was used. Finally, no significant difference (P ⫽ .695; two-tailed paired Student’s t-test) was found between the segmented, red ⫹ green pixel volume and the histologic estimation of viable neoplastic tissue. Unlike the aforementioned congruence, we observed a significant difference between the volume of red ⫹ green pixels and the histologically determined total volume of the residual tumor, as well as a low Pearson correlation of 0.74 (P ⬍ .0011) between these groups. A low Pearson correlation of 0.72 (P ⬍ .0017) was also found between the total volume of the red ⫹ green ⫹ blue pixels and the histologically determined total volume of residual tumor, but the volumes in the two groups were not significantly different (P ⬍ .135, two-tailed paired Student’s t-test).
565
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Figure 2. Subtracted and “three time point” (3TP) parametric images of localized infiltrating ductal carcinoma before and after neoadjuvant systemic therapy (NST) (Patient 8). (a) Subtracted (2 minutes postcontrast – zero time point) and 3TP parametric images of three central tumor slices before NST. (b) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from 3TP parametric images of the entire breast prior to NST. (c) Subtracted (2 minutes postcontrast minus zero time point) and parametric images of three central tumor slices after NST. (d) Frequency histograms of the intensity of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from 3TP parametric images of the entire breast after NST. Note:—Because of the applied threshold of 30% enhancement in the first postcontrast scan, and the scaling of the color intensity based on the calibration map, the color intensity range started from 104. *Indicates the position of the median value.
Based on these correlations, we estimated the effect of NST on cellular neoplastic volume by comparing changes in the volume of red ⫹ green pixels in each patient before and after treatment (Table 2). Most important, segmentation of the parametric images also enabled detection of very small (⬍1 mm) cancerous foci coded in red and green (Fig 5). We found a significant reduction (P ⬍ .0005; one-tailed paired Student’s t-test) in the presumed neoplastic tissue in 12 patients: 6 (of 7) with localized focal lesions and 6 (of 9) with diffuse, scattered lesions. No change in the volume of red ⫹ green pixels was observed in one patient; in three other patients, an increase in this volume was detected. Figure 6 demonstrates 3TP
566
parametric images before and after NST of a lesion that increased in size by 57(%). Segmentation analysis according to pixel color also indicated that most tumors exhibited a reduction in the volume of blue pixels (Table 3). These pixels reflect continuous signal enhancement over time, with no washout during the first 6 minutes. Blue pixels may indicate noninvolved fibroglandular tissue or postchemotherapy reparative fibrosis, as well as slowly perfused neoplastic tissue. No correlation (Pearson correlation coefficient ⫽ 0.21; P ⬍ .43) was found between the histologic volume of fibrosis (Table 1) and the volume obtained by segmentation of the blue pixels. Thus, although some percentage of
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 3. Subtracted and “three time point” (3TP) parametric images of diffused infiltrating ductal carcinoma before and after neoadjuvant systemic therapy (NST) (Patient 12). (a) Subtracted (2 minutes post contrast – zero time point) and 3TP parametric images of three central tumor slices before NST. (b) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from images of the entire breast before NST. (c) Subtracted (2 minutes postcontrast – zero time point) and 3TP parametric images of three central tumor slices after NST. (d) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawings) analyzed from images of the entire breast after NST. Note:—Because of the applied threshold of 30% enhancement in the first postcontrast scan, and the scaling of the color intensity based on the calibration map, the color intensity range started from 104.*Indicates the position of the median value.
blue pixels probably represents fibrosis, we cannot separate these pixels from other blue pixels indicating slowly perfused, noninvolved, or cancerous tissue. Moreover, some fibrosis within the residual tumor bed could be underestimated because of less than 30% enhancement in the first postcontrast scan. The parametric images were also used to scale the initial rate of enhancement, according to color intensity. Analysis of the intensities of the two groups of pixels (red ⫹ green; blue) showed that the former pixel group exhibited significantly higher color intensity, both before (P ⬍ .00028; twotailed paired Student’s t-test) and after (P ⬍ .023) NST
(Tables 2, 3). The higher intensity seen in the red ⫹ green pixels reflected high washin rates (high transcapillary transfer constants), as would be expected for viable neoplastic regions. Though changes in color intensity after NST varied among individual patients, the median percentile of the intensity in the red ⫹ green pixel group significantly decreased (P ⬍ .016; two-tailed paired Student’s t-test), whereas the change in intensity of the blue pixels was not statistically significant (P ⬍ .075; two-tailed paired Student’s t-test). Overall, a reduction in total tumor volume after NST was accompanied by a reduction in the intensity of the red and green pixels (Table 2).
567
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Table 2 Evaluation of the Segmentation and Intensity of the Regions Colored Red and Green in the Parametric Images Obtained Pre- and Post-NST Volume of Red ⫹ Green Pixels, cm3
Median Color Intensity of Red ⫹ Green Pixels
Case No.
Pretreatment
Posttreatment
Change (%)
Pretreatment
Posttreatment
Change (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
17.8 2.5 39.9 8.1 23.5 14.6 34.0 12.3 2.6 24.6 8.9 41.1 1.7 3.3 20.2 2.7 NA
0.1 0.2 4.0 1.0 4.9 3.9 11.6 4.9 1.1 11.8 4.7 28.8 1.7 3.5 31.7 4.4 17.9
⫺99 ⫺92 ⫺90 ⫺87 ⫺79 ⫺73 ⫺66 ⫺60 ⫺59 ⫺52 ⫺47 ⫺30 0 ⫹4 ⫹57 ⫹62 NA
140 160 213 186 110 210 235 239 194 216 126 192 207 152 177 158 NA
116 120 122 125 125 255 170 138 135 210 120 149 219 134 213 158 211
⫺17 ⫺25 ⫺43 ⫺33 ⫹14 ⫹21 ⫺28 ⫺42 ⫺30 ⫺3 ⫺5 ⫺22 ⫹6 ⫺12 ⫹20 0 NA
NST: neoadjuvant systemic therapy; NA: not available.
DISCUSSION The use of neoadjuvant systemic therapy is considered to be the standard treatment for the management of inoperable primary breast cancer. For operable breast cancer, it is employed as an alterative to adjuvant chemotherapy. In this pilot study, we demonstrated the use of a parametric imaging method based on DCE MRI to objectively assess the response of locally advanced breast cancer to NST, by evaluating changes induced in tumor localization, volume, and vascular function. In particular, we focused on developing a standardized approach involving segmentation of tumor tissue based on patterns of contrast enhancement and coded according to color hue and intensity, which does not depend on the radiologist’s delineation of tumor region of interest. This approach is of significant importance, because the effect of NST on breast tumors is frequently associated with tumor fragmentation to scattered foci that are impalpable in a clinical examination (34). Our approach also enabled the application of high spatial resolution at both the scanning and image processing stages. The high spatial resolution not only demonstrated the heterogeneity of contrast enhancement within each tumor, but also revealed the detailed changes in the enhancement patterns in response to NST.
568
Using the 3TP parametric method, we found that the combined red and green pixels, reflecting rapid washout of contrast material and hence, high efflux transfer constants, also displayed significantly higher color intensity than the blue pixels, indicating high influx transcapillary transfer constants. It was previously shown that cancerous breast tissue demonstrates increased influx and efflux transfer constants (32). This finding, suggesting that the green and red pixels represent regions with viable neoplastic cells, formed the basis for our segmentation method. Furthermore, this finding is in agreement with the results presented by El Khoury et al (13) and Partridge et al (14), and is based on the use of washout patterns for assessing response to NST. In a subsequent work by Partridge et al (15), it was also suggested that tumor size, as derived from the washout analysis before NST, is a predictor of recurrence-free survival. Moreover, NST-induced changes in the tumor size may provide a sensitive assessment of treatment efficacy. We were curious to investigate the origin of the enhanced pixels coded in blue. Although we did not find a correlation between the blue pixels and the fibrosis or necrosis (confirmed by histology), in some cases we could clearly identify in breast tissue from the same patient a change from red ⫹ green pixels before NST, to blue pixels after NST, suggesting that a transformation
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 4. Correlation of magnetic resonance imaging (MRI) and histopathologic analyses of viable tumor volume. (a) Pearson correlation and (b) Bland-Altman plot of the residual viable neoplastic volume determined by histologic grid morphometric analysis, as compared with the volume of the red ⫹ green pixels, determined by segmenting the “three time point” (3TP) parametric images from each patient. The graph displays a scatter diagram of the differences between the MRI and pathology derived tumor volumes plotted against the averages of these two measurements. Horizontal lines are drawn at the mean difference, and at the mean difference plus and minus the standard deviation of the differences. The error in estimating tumor volume from the 3TP – parametric images was estimated to be less than 10% of that volume and resulted from the nonautomated region of interest delineation of the whole breast which included enhanced regions outside the lesion (ie, blood vessels or normal breast parenchyma).
from cancerous cells to fibrosis had occurred. Development of algorithms that register the images before and after NST and depict, pixel-by-pixel, the spatial changes in the contrast dynamics, may help us to distinguish between noncancerous fibroglandular tissue, and repair fibrosis developed in response to treatment. Routine evaluation of mastectomy specimens provided us with an opportunity to compare histologic results with DCE-MRI data after chemotherapy. However, even patients with good clinical responses to NST could harbor innumerable small foci of residual invasive carcinoma scattered throughout the original tumor bed, which could be missed on gross histopathologic examination. This can occur because of the sectioning intervals, especially if the
lesion is not easily palpable after NST. Thus there are experimental impediments to correlating DCE-MRI findings with the results of routine pathologic evaluation of breast specimens. In this study, we therefore analyzed the posttreatment parametric images in a manner similar to the analysis of histologic specimens; namely, slice by slice, every 5 mm. This approach was particularly important in cases of diffuse disease with small, scattered foci, where the exact demarcation of residual tumor tissue is more difficult. However, because of the relatively small number of cases studied, the analyses of the parametric images, as well as data concerning disease prognosis and recurrence, are not amenable to statistical evaluation.
569
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Figure 5. “Three time point” parametric image of a central axial breast slice from a patient whose tumor responded to therapy (Patient 6). (a) Before neoadjuvant systemic therapy (NST). (b) After NST. Note the change in coloring (from red ⫹ green, to blue) and the presence of residual small cancerous loci (red) after NST.
Figure 6. “Three time point” parametric image of a central axial breast slice from a patient whose tumor failed to respond to therapy (Patient 15). (a) Before neoadjuvant systemic therapy (NST). (b) After NST. Note the substantial increase in tumor size and in area colored red after NST.
570
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Table 3 Evaluation of the Segmentation and Intensity of the Regions Colored Blue in the Parametric Images Obtained Pre- and Post-NST Volume of Blue Pixels, cm3
Median Color Intensity of Blue Pixels
Case No.
Pretreatment
Posttreatment
Change (%)
Pretreatment
Posttreatment
Change (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
6.4 1.6 16.0 5.1 5.9 4.0 13.4 5.2 1.9 11.7 5.6 19.2 1.1 4.4 14.9 2.2 NA
0.3 0.3 2.4 0.9 2.8 5.0 4.2 2.2 1.2 3.4 9.7 20.6 0.4 3.4 16.0 1.3 10.5
⫺96 ⫺81 ⫺85 ⫺83 ⫺52 ⫹26 ⫺69 ⫺57 ⫺38 ⫺71 ⫹75 ⫹7 ⫺64 ⫺23 ⫹7 ⫺42 NA
134 150 146 159 102 170 147 135 155 154 134 147 162 141 159 133 NA
120 120 123 126 127 178 141 135 136 140 129 141 158 131 159 147 158
⫺10 ⫺20 ⫺16 ⫺21 ⫹25 ⫹5 ⫺4 0 ⫺12 ⫺9 ⫺4 ⫺4 ⫺2 ⫺7 0 ⫹11 NA
NST: neoadjuvant systemic therapy; NA: not available.
It was previously shown that patients with significant tumor regression after NST experienced a longer period of disease-free survival than those whose tumors were less responsive to treatment (15). As expected from the generally known prognostic factors in cases of infiltrating ductal carcinoma (35,36), all four recurrences in our study occurred in patients with Stage III cancers: three of the four patients were diagnosed with lymph node metastasis, and the fourth had developed a very large residual mass (60.9 cm3). In contrast, we found that chemotherapeutic treatment of the three patients with lymph node metastasis induced appreciable regression in tumor volume by 92%, 66%, and 30%, respectively, as well as a reduction in the color intensity of the red and green pixels, indicating an appreciable response to therapy. Thus it appears that both regression in tumor size and suppression of vascular transcapillary transfer can be used as markers for monitoring response to treatment; however, these factors may not be as effective as lymph node metastasis in predicting recurrence of disease. Notably, the fourth recurrence of carcinoma after mastectomy was of a large tumor, within which, despite NST, both the number and color intensity of red plus green pixels increased. This may suggest that quantification of the changes in the 3TP parametric images may be more effective in predicting recurrence rates
in lymph node–negative than in lymph node–positive patients. In our parametric images, the washout pattern (color hue) is related to the efflux transcapillary transfer rate constant. According to the common physiologic models applied to analyze contrast enhanced breast MRI (37), this constant is, in turn, directly proportional to the outflux transcapillary transfer constant and inversely proportional to the extravascular, extracellular volume fraction. The estimated washin rate (denoted by color intensity) is related to the influx transcapillary constant. We assumed that the influx and outflux transfer constants are equal; hence, the method was standardized in terms of the transcapillary transfer constant and extravascular, extracellular volume fraction. This and other assumptions underlying the common physiologic models (37– 40), could prove to be invalid, preventing interpretation of the color scheme in terms of the transcapillary transfer constants. Furthermore, it is clear that, because of the limited number of time points chosen, this method cannot be used to determine the exact values of the transcapillary transfer constants. Rather, it can only provide approximate assessment of these constants (Fig 1), for use in the assembly and then segmentation of similar enhancement patterns, enabling discrimination between malignant and noninvolved
571
CHOU ET AL
breast tissue. Future development of DCE-MRI, with high spatial and temporal resolution and high signal-to-noise ratio, would make it possible to construct a more comprehensive physiologic model for quantifying changes in the values of the model’s various parameters. However, only large-scale clinical studies can verify whether changes in the parameters of a more comprehensive model are, indeed, critical predictors of tumor response to therapy, and better indicators of possible disease recurrence than the approximated parameters described inhere. In summary, we have demonstrated the application of a standardized, high-resolution parametric dynamic contrast-enhanced magnetic resonance imaging method, to monitor and assess the response of breast cancer to NST. This method provides an objective means by which to segment the volume of pixels reflecting viable tumor tissue, both before and after neoadjuvant systemic therapy, and to estimate in the course of treatment changes in this volume, as well as in the vascular transcapillary transfer properties. Additional MRI data from large-scale clinical studies, for comparison with prognostic factors obtained by histopathologic and immunohistochemical assessment of tumor tissue as well as long-term follow-up of patients are necessary to validate the clinical efficacy of this method to help predict recurrence/survival rates. ACKNOWLEDGMENTS
We would like to thank Dr. Dov Grobgeld for the development of the software and Ms. Barbara Morgenstern for skillful editing. H. Degani is the incumbent of the Fred and Andrea Fallek Professorial Chair for Breast Cancer Research. REFERENCES 1. Kaufmann M, Hortobagyi GN, Goldhirsch A, et al. Recommendations from an international expert panel on the use of neoadjuvant (primary) systemic treatment of operable breast cancer: an update. J Clin Oncol 2006; 24:1940 –1949. 2. Helbich TH. Contrast-enhanced magnetic resonance imaging of the breast. Eur J Radiol 2000; 34:208 –219. 3. Padhani AR, Ah-See ML, Makris A. MRI in the detection and management of breast cancer. Expert Rev Anticancer Ther 2005; 5:239 –252. 4. Kaiser WA, Zeitler E. MR imaging of the breast: fast imaging sequences with and without Gd-DTPA. Preliminary observations. Radiology 1989; 170:681– 686. 5. Knopp MV, von Tengg-Kobligk H, Choyke PL. Functional magnetic resonance imaging in oncology for diagnosis and therapy monitoring. Mol Cancer Ther 2003; 2:419 – 426. 6. Schnall MD, Blume J, Bluemke DA, et al. Diagnostic architectural and dynamic features at breast MR imaging: multicenter study. Radiology 2006; 238:42–53. 7. Rieber A, Brambs HJ, Gabelmann A, et al. Breast MRI for monitoring response of primary breast cancer to neo-adjuvant chemotherapy. Eur Radiol 2002; 12:1711–1719.
572
Academic Radiology, Vol 14, No 5, May 2007
8. Wasser K, Sinn HP, Fink C, et al. Accuracy of tumor size measurement in breast cancer using MRI is influenced by histological regression induced by neoadjuvant chemotherapy. Eur Radiol 2003; 13:1213–1223. 9. Esserman L, Kaplan E, Partridge S, et al. MRI phenotype is associated with response to doxorubicin and cyclophosphamide neoadjuvant chemotherapy in stage III breast cancer. Ann Surg Oncol 2001; 8:549 –559. 10. Cheung YC, Chen SC, Su MY, et al. Monitoring the size and response of locally advanced breast cancers to neoadjuvant chemotherapy (weekly paclitaxel and epirubicin) with serial enhanced MRI. Breast Cancer Res Treat 2003; 78:51–58. 11. Rosen EL, Blackwell KL, Baker JA, et al. Accuracy of MRI in the detection of residual breast cancer after neoadjuvant chemotherapy. AJR Am J Roentgenol 2003; 181:1275–1282. 12. Martincich L, Montemurro F, De Rosa G, et al. Monitoring response to primary chemotherapy in breast cancer using dynamic contrast-enhanced magnetic resonance imaging. Breast Cancer Res Treat 2004; 83:67–76. 13. El Khoury C, Servois V, Thibault F, et al. MR quantification of the washout changes in breast tumors under preoperative chemotherapy: feasibility and preliminary results. AJR Am J Roentgenol 2005; 184:1499 –1504. 14. Partridge SC, Gibbs JE, Lu Y, et al. Accuracy of MR imaging for revealing residual breast cancer in patients who have undergone neoadjuvant chemotherapy. AJR Am J Roentgenol 2002; 179:1193–1199. 15. Partridge SC, Gibbs JE, Lu Y, et al. MRI measurements of breast tumor volume predict response to neoadjuvant chemotherapy and recurrence-free survival. AJR Am J Roentgenol 2005; 184:1774 –1781. 16. Delille JP, Slanetz PJ, Yeh ED, et al. Invasive ductal breast carcinoma response to neoadjuvant chemotherapy: noninvasive monitoring with functional MR imaging pilot study. Radiology 2003; 228:63– 69. 17. Chang YC, Huang CS, Liu YJ, et al. Angiogenic response of locally advanced breast cancer to neoadjuvant chemotherapy evaluated with parametric histogram from dynamic contrast-enhanced MRI. Phys Med Biol 2004; 49:3593–3602. 18. Armitage P, Behrenbruch C, Brady M, et al. Extracting and visualizing physiological parameters using dynamic contrast-enhanced magnetic resonance imaging of the breast. Med Image Anal 2005; 9:315–329. 19. Padhani AR, Hayes C, Assersohn L, et al. Prediction of clinicopathologic response of breast cancer to primary chemotherapy at contrastenhanced MR imaging: initial clinical results. Radiology 2006; 239:361– 374. 20. Pickles MD, Lowry M, Manton DJ, et al. Role of dynamic contrast enhanced MRI in monitoring early response of locally advanced breast cancer to neoadjuvant chemotherapy. Breast Cancer Res Treat 2005; 91:1–10. 21. Hayes C, Padhani AR, Leach MO. Assessing changes in tumour vascular function using dynamic contrast-enhanced magnetic resonance imaging. NMR Biomed 2002; 15:154 –163. 22. Manton DJ, Chaturvedi A, Hubbard A, et al. Neoadjuvant chemotherapy in breast cancer: early response prediction with quantitative MR imaging and spectroscopy. Br J Cancer 2006; 94:427– 435. 23. Fischer U, Zachariae O, Baum F, et al. The influence of preoperative MRI of the breasts on recurrence rate in patients with breast cancer. Eur Radiol 2004; 14:1725–1731. 24. Thibault F, Nos C, Meunier M, et al. MRI for surgical planning in patients with breast cancer who undergo preoperative chemotherapy. AJR Am J Roentgenol 2004; 183:1159 –1168. 25. Denis F, Desbiez-Bourcier AV, Chapiron C, et al. Contrast enhanced magnetic resonance imaging underestimates residual disease following neoadjuvant docetaxel based chemotherapy for breast cancer. Eur J Surg Oncol 2004; 30:1069 –1076. 26. Degani H, Gusis V, Weinstein D, et al. Mapping pathophysiological features of breast tumors by MRI at high spatial resolution. Nat Med 1997; 3:780 –782. 27. Furman-Haran E, Degani H. Parametric analysis of breast MRI. J Comput Assist Tomogr 2002; 26:376 –386. 28. Kelcz F, Furman-Haran E, Grobgeld D, et al. Clinical testing of highspatial-resolution parametric contrast-enhanced MR imaging of the breast. AJR Am J Roentgenol 2002; 179:1485–1492.
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
29. Furman-Haran E, Grobgeld D, Margalit R, et al. Response of MCF7 human breast cancer to tamoxifen: evaluation by the three-time-point, contrast-enhanced magnetic resonance imaging method. Clin Cancer Res 1998; 4:2299 –2304. 30. Tofts PS, Kermode AG. Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 1991; 17:357–367. 31. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. J Magn Reson Imaging 1997; 7:91–101. 32. Furman-Haran E, Schechtman E, Kelcz F, et al. Magnetic resonance imaging reveals functional diversity of the vasculature in benign and malignant breast lesions. Cancer 2005; 104:708 –718. 33. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000; 92:205–216. 34. Moneer M, El-Didi M, Khaled H. Breast conservative surgery: is it appropriate for locally advanced breast cancer following downstaging by neoadjuvant chemotherapy? A pathological assessment. Breast 1999; 8:315–319.
35. Singletary SE, Connolly JL. Breast cancer staging: working with the sixth edition of the AJCC Cancer Staging Manual. CA Cancer J Clin 2006; 56:37– 47; quiz 50 –31. 36. Yarbro JW, Page DL, Fielding LP, et al. American Joint Committee on Cancer Prognostic Factors Consensus Conference. Cancer 1999; 86: 2436 –2446. 37. Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters from dynamic contrast-enhanced T1-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10:223–232. 38. Yankeelov TE, Rooney WD, Li X, et al. Variation of the relaxographic “shutter-speed” for transcytolemmal water exchange affects the CR bolus-tracking curve shape. Magn Reson Med 2003; 50:1151–1169. 39. Dadiani M, Margalit R, Sela N, et al. High-resolution magnetic resonance imaging of disparities in the transcapillary transfer rates in orthotopically inoculated invasive breast tumors. Cancer Res 2004; 64:3155– 3161. 40. Hassid Y, Furman-Haran E, Margalit R, et al. Non invasive magnetic resonance imaging of transport and interstitial fluid pressure in ectopic human lung tumors. Cancer Res 2006; 66:4159.
573
Rationale and Objectives. Neoadjuvant systemic therapy (NST) is the standard treatment for locally advanced breast cancer and a common option for primary operable disease. It is important to develop standardized imaging techniques that can monitor and quantify response to NST enabling treatment tailored to each individual patient, and facilitating surgical planning. Here we present a high spatial resolution, parametric method based on dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI), which evaluates breast cancer response to NST. Materials and Methods. DCE-MRI examinations were performed twice on 17 breast cancer patients, before and after treatment. Seven sets of axial breast images were sequentially recorded at 1.5 Tesla applying a three-dimensional, gradient echo at a spatial resolution ⬃2 ⫻ 1.2 ⫻ 0.6 mm3 and temporal resolution ⬃2 minutes, using gadopentate dimeglumine (0.1 mmol/kg wt). Image analysis was based on a color-coded scheme related to physiologic perfusion parameters. Results. A high Pearson correlation coefficient of 0.96 (P ⬍ .0001) was found between the histopathologic estimation of viable neoplastic tissue volume and the segmented volume of all the pixels demonstrating fast and steady state washout after NST (colored in light red and green). Segmentation of these pixels before and after NST indicated response in terms of reduced tumor volume and a parallel decrease in enhancement rate which reflects diminished transcapillary transfer of the contrast agent. Conclusions. The use of a parametric MRI technique provided a means to standardize segmentation and quantify changes in the perfusion of breast neoplastic tissue in response to NST. Whether this technique can serve to predict breast cancer recurrence and survival rates requires further clinical testing. Key Words. Breast cancer; MRI; neoadjuvant systemic therapy; prognosis. ©
AUR, 2007
Acad Radiol 2007; 14:561–573 1
From the Departments of Radiology (C.-P.C., M.-T.W., H.-B.P.), Surgery (H.T.C.), and Pathology and Laboratory Medicine (Y.-S.L.), Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ROC; School of Medicine, National Yang Ming University, Taipei, Taiwan, ROC (C.-P.C., M.-T.W., H.-T.C., Y.-S.L., H.-B.P.); Department of Biological Regulation, Weizmann Institute of Science, Rehovot 76100, Israel (H.D., E.F.-H.). Received January 27, 2007; accepted February 7, 2007. Supported by a research grant from Kaohsiung Veterans General Hospital, VGHKS-91– 84, an Israel Science Foundation grant (number 801/04) and by Lord David Alliance, United Kingdom. Address correspondence to: E.F.H. e-mail: edna.haran@ weizmann.ac.il
© AUR, 2007 doi:10.1016/j.acra.2007.02.005
Neoadjuvant systemic therapy (NST) is the standard treatment for locally advanced breast cancer and a common option for primary operable disease. It is mainly designed to reduce tumor size, thereby improving surgical outcomes; to evaluate response to systemic therapy; and to obtain long-term, disease-free survival (1). To achieve these goals, it is necessary to develop standardized imaging techniques that can noninvasively monitor the response of breast tumors to NST and quantify changes in their size and spread, as well as track specific biologic and physiologic markers of malignancy. Such imaging techniques could be useful in the early stages of treatment, to help predict response to chemotherapy, to enable
561
CHOU ET AL
treatment tailored to each individual patient, and to facilitate surgical planning at later treatment stages. Magnetic resonance imaging (MRI), an important adjuvant tool for the detection and characterization of breast cancer (2– 4) has also been used to monitor the effects of NST (5). Specifically, contrast-enhanced MRI provides a means to delineate the architectural and dynamic features of breast tumors and determine their size (6). In recent years, several groups have examined whether contrastenhanced MRI is accurate enough to evaluate residual tumor size after NST, and compared their results with final histopathologic assessment of mastectomy specimens (7–11) and with data from other imaging methods such as mammography and sonography. In addition to detecting variations in tumor size, dynamic contrast-enhanced (DCE) MRI can provide information regarding the pathophysiologic response of the tumor vasculature to NST. It is well-known that tumor angiogenesis leads to the formation of blood microvessels that are excessively permeable and enhance leakage of bloodborne contrast agents, leading to augmented contrast enhancement. In tumors responding to chemotherapy, tumor angiogenesis is halted, and the development of necrosis and fibrosis leads to the establishment of a microcapillary network with properties different from that feeding the growing tumor. These changes can be quantified by analyzing the enhancement parameters of dynamic contrast-enhanced images (12–15) or by using tracer pharmacokinetic models that calculate transcapillary transfer constants and other related parameters (8,16 –18). Modelbased DCE analysis both before and after the first or second chemotherapeutic treatment has also proven useful in predicting its final outcome, and may aid in designing the type and timing of therapy (19 –22). Thus, in addition to changes in tumor size, variations in the transcapillary transfer constants can help to identify patients likely to be nonresponsive to a particular therapy, early in treatment. Although some studies have suggested that MRI evaluation before breast surgery can decrease local recurrence rates after breast conservation (23), the potential to clarify the clinical picture for patients undergoing NST, still has some controversial issues (24,25). Because tumors are usually heterogeneous in nature, it is important to evaluate their response to neoadjuvant systemic therapy at high spatial resolution. However, in practical terms, it is difficult to achieve images with sufficient signal-to-noise at the temporal resolution dictated by the necessity for high spatial resolution to optimize the precision of the parameters derived from the fitting procedures
562
Academic Radiology, Vol 14, No 5, May 2007
required for pharmacokinetic modeling. In this article, we present a pilot study demonstrating the use of a high-resolution, parametric method based on DCE MRI to evaluate response to NST. This method was originally developed to improve breast cancer diagnosis (26,27) and, in that context, demonstrated high sensitivity and specificity in differentiating between benign and malignant breast lesions (28). It was also employed in a mouse model of human breast cancer xenografts to monitor response to hormonal therapy (29). In the present study, we demonstrate that this technique enables the design of an objective segmentation procedure for quantifying changes in tumor volume and the spread of the disease in the breast, as well as estimating variations in microvascular function within viable neoplastic regions.
MATERIALS AND METHODS Patients, Treatment, and Timing of MRI A total of 17 patients, 35– 62 years of age (median age, 48 years) were recruited to this study, which was carried out between 2001 and 2003. All of the patients signed an informed consent form approved by the institutional review board of the hospital. Patients were initially diagnosed with Stage II-III infiltrating ductal carcinoma, using incision biopsy or core needle biopsy, and were scheduled for NST. Other requirements for entry into the study included patient age (⬍70), lack of pregnancy, and demonstration of adequate hematologic, renal, hepatic, and cardiac function. Patients received two to four courses of treatment with 5-fluorouracil (500 mg/m2), epi-Adriamycin (70 mg/m2), and cyclophosphamide (500 mg/m2), at 3-week intervals. Breast MRI scans were performed twice, before NST commenced and after it concluded, before surgery. In one case, a technical failure precluded analysis of the images recorded before the start of chemotherapy. Histologic Evaluation After NST, patients underwent a mastectomy and then, the breast specimens (n ⫽ 16) were sectioned at 5-mm increments, perpendicular to the cranial-caudal axis, and stained with hematoxylin and eosin for histopathologic evaluation. This evaluation was conducted by the same experienced pathologist with no prior knowledge of preoperative MRI findings. In one case, the breast specimen was not evaluated systematically in this way and hence was not included in the correlation analysis with the MRI
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 1. Color-coded calibration map of the “three time point” (3TP) parametric images. (a) A color-coding scheme of the various enhancement patterns using selected imaging sets at three time points. The washout pattern is coded by color hue: red signifies a decline in signal intensity from the first to the second postcontrast time points; green, no change in signal intensity in the first and second postcontrast time points, within an average noise level; and blue, an increase in signal intensity from the first to the second postcontrast time points. The washin rate is denoted by color intensity (in arbitrary units ranging from 0 to 256) and normalized to the maximum intensity in the calibration map. (b) A calibration map calculated using the experimental conditions of this study. (c) Simulated enhancement curves of the white dots in (b) for (ktrans,ve) ⫽ (0.9, 0.6) in red and (ktrans,ve) ⫽ (0.2, 0.4) in blue. The broken black lines in (c) demonstrate the deviation of the estimated initial rates determined by the 3TP software, from the initial rates simulated on the basis of the Tofts model (31).
findings. Pathologic evaluations of various tissue volumes in the residual tumor bed, and, within each volume, the fraction of viable neoplastic cells, fibrosis, and necrosis, were performed using a grid morphometric analysis. The fractions of the various tissue types within the residual mass also included scattered microscopic cancerous foci. MRI All of the examinations were performed using a 1.5Tesla MRI scanner equipped with a two-channel, phased array breast coil (GE Medical Systems, Waukesha, WI). MRI of both breasts were acquired in the axial plane by using a three-dimensional gradient echo sequence without fat suppression with echo time/repetition time ⫽ 4.2/7.9 milliseconds; flip angle ⫽ 30°; field of view ⫽ 32– 40 cm; matrix ⫽ 64 ⫻ 256 ⫻ 512 interpolated to 120 ⫻ 512 ⫻ 512, with a final slice thickness of 1.0 mm. Seven image sets, each taken over approximately 2 minutes, were sequentially recorded. The contrast agent, gadopentate dimeglumine (Magnevist, Schering, Berlin, Germany), was injected intravenously in the middle of the second image set by means of an automated injection system (Optistar MR injector, Mallinckrodt, Hazelwood, MO), using a dose of 0.1 mmol/kg body weight at a flow rate of 3 mL/second, followed by a 10-mL saline flush. Image Analysis Images were transferred to a Linux-operated personal computer and processed at pixel resolution using in-house
developed algorithms, as well as the “three time point” (3TP) pharmacokinetic software (26). This software is based on the Tofts physiologic model (30,31) that relates the time evolution of contrast enhancement in a tissue, to the influx transcapillary transfer constant, kin, and the efflux transcapillary transfer constant, kep. In turn, kep is equal to the ratio of the outward transcapillary transfer constant, and the extravascular, extracellular volume fraction, kout/ve. The 3TP software selects three sets of images, the first recorded before injection of the contrast agent (corresponding to time point zero) and the second and third recorded postcontrast (corresponding to the first and second time points, postcontrast). Each pixel was then coded by means of color hue and color intensity (Fig 1). Color hue reflects changes in signal intensity between the second and third time points, and is related to the washout pattern of the contrast agent: fast clearance is coded red; steady-state clearance coded green; and continuous entrance of the contrast agent with no clearance, coded blue. Each color hue is presented in arbitrary units ranging from 0 to 256, and normalized to the maximum intensity in a calibration map (see next paragraph), reflecting the rate of change in signal intensity between the first and second time points (apparent initial rate) (Fig 1). The execution time of the 3TP algorithm on a standard P4 2.8 GHz PC was approximately 30 seconds for each case. To achieve standardized utilization of this color-coded scheme, it was necessary to select two specific postcontrast time points. This selection is based on construction
563
CHOU ET AL
of a calibration map that simulates the Tofts model (30,31), and takes into account the specific set of the imaging parameters recorded (type of sequence, echo time, repetition time, flip angle), T1 and T2 relaxivities, the plasma pharmacokinetic parameters of the contrast agent, and the T1 relaxation rate of the imaged tissue, which depends on the magnetic field strength of the scanner. In building the calibration map, we assumed that the influx and outflux transcapillary transfer constants are equal; hence, the y-axis corresponds to ktrans (kin ⫽ kout ⫽ ktrans) and the x-axis to ve (Fig 1b). For any two selected postcontrast time points and for each pair of ktrans and ve values, the software algorithm calculates theoretical signal intensities, and assigns a color hue and color intensity according to the scheme depicted in Fig 1a, yielding a color-coded calibration map standardized for the sequence parameters and contrast agent (Fig 1b). Thus the combination of color hue and color intensity in a parametric image reflects a specific combination of ktrans and ve values in the calibration map (Fig 1b), which determines the changes in concentration of the contrast agent over time (Fig 1c). Based on previous estimations of ktrans and ve values in infiltrating breast cancer (32) and on the contrast agent and field strength employed in our study, the following three time points were chosen for all parametric analyses: precontrast (time point zero), 2 minutes postcontrast, and 6 minutes postcontrast. To exclude normal glandular enhancement from the 3TP parametric analysis of the neoplastic tissue, we initially selected all the pixels that exhibited more than 30% enhancement in the first postcontrast scan. Using this threshold selection criterion and normalizing the maximal intensity by means of the calibration map yielded an arbitrary color intensity scale ranging from 104 to 256. We then segmented the parametric images by drawing a region of interest around the whole breast in every fifth consecutive slice (at intervals of 5 mm) similar to the grid morphometric analysis. In each case, we quantified the number of pixels exhibiting each color hue, both before and after chemotherapy. This enabled us to assess changes in the number of pixels of each color hue and hence changes in the volume of each pixel group induced by chemotherapy. The assessment of viable neoplasm before and after NST was based on the change in volume of red ⫹ green pixels on the 3TP parametric images. The response to NST was assessed in each patient according to the published Response Evaluation Criteria in Solid Tumors (33). We also generated distribution histograms
564
Academic Radiology, Vol 14, No 5, May 2007
of color intensity for each pixel group, and calculated the median value of this parameter, both before and after chemotherapy. Statistical Evaluation Correlation between pathology data and the analyses of the 3TP parametric images according to color hue was assessed using the Pearson correlation test and two-tailed, paired Student’s t-test. A Bland-Altman plot was employed to further evaluate whether the parameters obtained from analysis of the parametric images either underestimated or overestimated the corresponding histologically determined parameters. The paired Student’s t-test was also applied to evaluate chemotherapy-induced changes in color hue and color intensity of segmented pixels in the 3TP parametric images. All statistical analyses were carried out using SAS software (version 8.0 for Windows).
RESULTS Our objectives in this pilot study were twofold: to extend the 3TP parametric method, based on high-resolution, contrast-enhanced MRI to monitor the response of malignant breast tumors to neoadjuvant systemic therapy and to quantify, in a standardized manner, the changes it induced. Accordingly, 17 patients diagnosed with infiltrating ductal breast cancer and scheduled to receive NST were scanned twice, before and after treatment, and before mastectomy. Detailed histopathologic evaluation of the entire breast in axial planes, similar to those scanned by MRI, was conducted after mastectomy in 16 patients (1 patient underwent surgery and final histologic evaluation at another hospital). Of these 16 cases, 15 were diagnosed with infiltrating ductal carcinoma, and 1 with a mixture of infiltrating ductal carcinoma, and ductal carcinoma in situ. None of the tumors showed complete pathologic response to the treatment. For each patient, the histopathologic features of mastectomy specimens and prognostic factors and the response to NST and disease recurrence over time are summarized in Table 1. Grid morphometric analysis yielded an estimate of the volume of the entire lesion postchemotherapy, as well as the percentage from this volume of viable neoplastic regions, fibrosis, and necrosis (Table 1). Before NST, the postcontrast images of the diseased breasts revealed marked signal enhancement, compared to
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Academic Radiology, Vol 14, No 5, May 2007
Table 1 Data from Mastectomy Specimens, Lymph Node Status, and Follow up % Viable NST Final Residual Recurrence-free Neoplastic Case No. Morphologic Pathologic Lymph Node Survival, Cells† No. Stage Cycle Distribution Volume† (cm3) % Fibrosis† % Necrosis† Response Involved Months 1 2 3 4 5 6 7 8 9 10 11 12 13 14* 15 16 17
III III II III III III III II II III II III III II III III II
3 3 4 3 3 3 3 3 2 3 3 4 2 2 3 3 3
Localized Localized Localized Diffused Localized Localized Localized Localized Diffused Diffused Localized Diffused Localized Diffused Diffused Diffused Diffused
NA 1.2 4.6 5.5 43.3 5.3 27.9 6.5 2.8 14.7 2.5 47.6 4.4 46.1 60.9 3.2 155
NA 4 59 22 15 47 70 62 47 76 25 62 78 5 47 94 10
NA 95 25 75 80 50 25 34 50 20 62 33 20 75 20 5 50
NA 1 16 3 5 3 5 4 3 4 13 5 2 20 33 1 40
PR PR PR PR PR PR PR SD SD SD SD SD SD SD SD SD SD
0 5 0 0 6 0 3 0 0 5 0 22 0 0 0 2 1
45 22‡ 33 44 39 50 9‡ 50 54 38 44 6‡ 44 38 12‡ 53 48
NST: neoadjuvant systemic therapy; NA: not available; PR: partial response; SD: stable disease (defined using Response Evaluation Criteria in Solid Tumors). *Mixed diagnosis: infiltrating ductal carcinoma and ductal carcinoma in situ. †Derived from grid morphometric analysis of residual tumor mass. ‡Recurrence.
the precontrast images (Fig 2a, first row, and Fig 3a, first row). After NST, enhancement patterns differed among patients, showing either localized (Fig 2c, first row) or scattered changes (Fig 3c, first row). Application of the 3TP software yielded color-coded parametric images (Fig 2a,c, second row; Fig 3a,c, second row) enabling the automated segmentation of pixels over a region of interest of the entire breast, according to their color hue and intensity. In all patients, the 3TP parametric images obtained before NST were characterized by a large fraction of red and green pixels, reflecting a fast or steady-state washout from high efflux transcapillary transfer rates and a small fraction of blue pixels, reflecting delayed washout from low efflux transcapillary transfer rates (Fig 2, 3). After NST, the total number of enhanced pixels, and hence colored pixels (red ⫹ green ⫹ blue), changed in each patient. The large volume fraction of red and green pixels in all lesions, as well as previous studies indicating the utility of a washout pattern for identifying tumor tissue (13,14) suggested that these pixels could be used as markers for segmenting viable tumor regions. We tested this hypothesis by correlating the color segmentation results of the red and green pixels (Table 2) with the histo-
logic evaluations after NST (Table 1). This analysis indicated a high Pearson correlation coefficient of 0.96 (P ⬍ .0001) between the combined volumes of red ⫹ green pixels after NST and the pathologic estimation of cellular neoplastic volume, as shown in Fig 4a. Furthermore, a Bland-Altman plot (Fig 4b) indicated no bias in the direction of either over- or underestimation of the volume of viable tumor tissue, when the color segmentation method was used. Finally, no significant difference (P ⫽ .695; two-tailed paired Student’s t-test) was found between the segmented, red ⫹ green pixel volume and the histologic estimation of viable neoplastic tissue. Unlike the aforementioned congruence, we observed a significant difference between the volume of red ⫹ green pixels and the histologically determined total volume of the residual tumor, as well as a low Pearson correlation of 0.74 (P ⬍ .0011) between these groups. A low Pearson correlation of 0.72 (P ⬍ .0017) was also found between the total volume of the red ⫹ green ⫹ blue pixels and the histologically determined total volume of residual tumor, but the volumes in the two groups were not significantly different (P ⬍ .135, two-tailed paired Student’s t-test).
565
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Figure 2. Subtracted and “three time point” (3TP) parametric images of localized infiltrating ductal carcinoma before and after neoadjuvant systemic therapy (NST) (Patient 8). (a) Subtracted (2 minutes postcontrast – zero time point) and 3TP parametric images of three central tumor slices before NST. (b) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from 3TP parametric images of the entire breast prior to NST. (c) Subtracted (2 minutes postcontrast minus zero time point) and parametric images of three central tumor slices after NST. (d) Frequency histograms of the intensity of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from 3TP parametric images of the entire breast after NST. Note:—Because of the applied threshold of 30% enhancement in the first postcontrast scan, and the scaling of the color intensity based on the calibration map, the color intensity range started from 104. *Indicates the position of the median value.
Based on these correlations, we estimated the effect of NST on cellular neoplastic volume by comparing changes in the volume of red ⫹ green pixels in each patient before and after treatment (Table 2). Most important, segmentation of the parametric images also enabled detection of very small (⬍1 mm) cancerous foci coded in red and green (Fig 5). We found a significant reduction (P ⬍ .0005; one-tailed paired Student’s t-test) in the presumed neoplastic tissue in 12 patients: 6 (of 7) with localized focal lesions and 6 (of 9) with diffuse, scattered lesions. No change in the volume of red ⫹ green pixels was observed in one patient; in three other patients, an increase in this volume was detected. Figure 6 demonstrates 3TP
566
parametric images before and after NST of a lesion that increased in size by 57(%). Segmentation analysis according to pixel color also indicated that most tumors exhibited a reduction in the volume of blue pixels (Table 3). These pixels reflect continuous signal enhancement over time, with no washout during the first 6 minutes. Blue pixels may indicate noninvolved fibroglandular tissue or postchemotherapy reparative fibrosis, as well as slowly perfused neoplastic tissue. No correlation (Pearson correlation coefficient ⫽ 0.21; P ⬍ .43) was found between the histologic volume of fibrosis (Table 1) and the volume obtained by segmentation of the blue pixels. Thus, although some percentage of
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 3. Subtracted and “three time point” (3TP) parametric images of diffused infiltrating ductal carcinoma before and after neoadjuvant systemic therapy (NST) (Patient 12). (a) Subtracted (2 minutes post contrast – zero time point) and 3TP parametric images of three central tumor slices before NST. (b) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawing) analyzed from images of the entire breast before NST. (c) Subtracted (2 minutes postcontrast – zero time point) and 3TP parametric images of three central tumor slices after NST. (d) Frequency histograms of the intensities of red ⫹ green pixels (upper drawing) and blue pixels (lower drawings) analyzed from images of the entire breast after NST. Note:—Because of the applied threshold of 30% enhancement in the first postcontrast scan, and the scaling of the color intensity based on the calibration map, the color intensity range started from 104.*Indicates the position of the median value.
blue pixels probably represents fibrosis, we cannot separate these pixels from other blue pixels indicating slowly perfused, noninvolved, or cancerous tissue. Moreover, some fibrosis within the residual tumor bed could be underestimated because of less than 30% enhancement in the first postcontrast scan. The parametric images were also used to scale the initial rate of enhancement, according to color intensity. Analysis of the intensities of the two groups of pixels (red ⫹ green; blue) showed that the former pixel group exhibited significantly higher color intensity, both before (P ⬍ .00028; twotailed paired Student’s t-test) and after (P ⬍ .023) NST
(Tables 2, 3). The higher intensity seen in the red ⫹ green pixels reflected high washin rates (high transcapillary transfer constants), as would be expected for viable neoplastic regions. Though changes in color intensity after NST varied among individual patients, the median percentile of the intensity in the red ⫹ green pixel group significantly decreased (P ⬍ .016; two-tailed paired Student’s t-test), whereas the change in intensity of the blue pixels was not statistically significant (P ⬍ .075; two-tailed paired Student’s t-test). Overall, a reduction in total tumor volume after NST was accompanied by a reduction in the intensity of the red and green pixels (Table 2).
567
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Table 2 Evaluation of the Segmentation and Intensity of the Regions Colored Red and Green in the Parametric Images Obtained Pre- and Post-NST Volume of Red ⫹ Green Pixels, cm3
Median Color Intensity of Red ⫹ Green Pixels
Case No.
Pretreatment
Posttreatment
Change (%)
Pretreatment
Posttreatment
Change (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
17.8 2.5 39.9 8.1 23.5 14.6 34.0 12.3 2.6 24.6 8.9 41.1 1.7 3.3 20.2 2.7 NA
0.1 0.2 4.0 1.0 4.9 3.9 11.6 4.9 1.1 11.8 4.7 28.8 1.7 3.5 31.7 4.4 17.9
⫺99 ⫺92 ⫺90 ⫺87 ⫺79 ⫺73 ⫺66 ⫺60 ⫺59 ⫺52 ⫺47 ⫺30 0 ⫹4 ⫹57 ⫹62 NA
140 160 213 186 110 210 235 239 194 216 126 192 207 152 177 158 NA
116 120 122 125 125 255 170 138 135 210 120 149 219 134 213 158 211
⫺17 ⫺25 ⫺43 ⫺33 ⫹14 ⫹21 ⫺28 ⫺42 ⫺30 ⫺3 ⫺5 ⫺22 ⫹6 ⫺12 ⫹20 0 NA
NST: neoadjuvant systemic therapy; NA: not available.
DISCUSSION The use of neoadjuvant systemic therapy is considered to be the standard treatment for the management of inoperable primary breast cancer. For operable breast cancer, it is employed as an alterative to adjuvant chemotherapy. In this pilot study, we demonstrated the use of a parametric imaging method based on DCE MRI to objectively assess the response of locally advanced breast cancer to NST, by evaluating changes induced in tumor localization, volume, and vascular function. In particular, we focused on developing a standardized approach involving segmentation of tumor tissue based on patterns of contrast enhancement and coded according to color hue and intensity, which does not depend on the radiologist’s delineation of tumor region of interest. This approach is of significant importance, because the effect of NST on breast tumors is frequently associated with tumor fragmentation to scattered foci that are impalpable in a clinical examination (34). Our approach also enabled the application of high spatial resolution at both the scanning and image processing stages. The high spatial resolution not only demonstrated the heterogeneity of contrast enhancement within each tumor, but also revealed the detailed changes in the enhancement patterns in response to NST.
568
Using the 3TP parametric method, we found that the combined red and green pixels, reflecting rapid washout of contrast material and hence, high efflux transfer constants, also displayed significantly higher color intensity than the blue pixels, indicating high influx transcapillary transfer constants. It was previously shown that cancerous breast tissue demonstrates increased influx and efflux transfer constants (32). This finding, suggesting that the green and red pixels represent regions with viable neoplastic cells, formed the basis for our segmentation method. Furthermore, this finding is in agreement with the results presented by El Khoury et al (13) and Partridge et al (14), and is based on the use of washout patterns for assessing response to NST. In a subsequent work by Partridge et al (15), it was also suggested that tumor size, as derived from the washout analysis before NST, is a predictor of recurrence-free survival. Moreover, NST-induced changes in the tumor size may provide a sensitive assessment of treatment efficacy. We were curious to investigate the origin of the enhanced pixels coded in blue. Although we did not find a correlation between the blue pixels and the fibrosis or necrosis (confirmed by histology), in some cases we could clearly identify in breast tissue from the same patient a change from red ⫹ green pixels before NST, to blue pixels after NST, suggesting that a transformation
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Figure 4. Correlation of magnetic resonance imaging (MRI) and histopathologic analyses of viable tumor volume. (a) Pearson correlation and (b) Bland-Altman plot of the residual viable neoplastic volume determined by histologic grid morphometric analysis, as compared with the volume of the red ⫹ green pixels, determined by segmenting the “three time point” (3TP) parametric images from each patient. The graph displays a scatter diagram of the differences between the MRI and pathology derived tumor volumes plotted against the averages of these two measurements. Horizontal lines are drawn at the mean difference, and at the mean difference plus and minus the standard deviation of the differences. The error in estimating tumor volume from the 3TP – parametric images was estimated to be less than 10% of that volume and resulted from the nonautomated region of interest delineation of the whole breast which included enhanced regions outside the lesion (ie, blood vessels or normal breast parenchyma).
from cancerous cells to fibrosis had occurred. Development of algorithms that register the images before and after NST and depict, pixel-by-pixel, the spatial changes in the contrast dynamics, may help us to distinguish between noncancerous fibroglandular tissue, and repair fibrosis developed in response to treatment. Routine evaluation of mastectomy specimens provided us with an opportunity to compare histologic results with DCE-MRI data after chemotherapy. However, even patients with good clinical responses to NST could harbor innumerable small foci of residual invasive carcinoma scattered throughout the original tumor bed, which could be missed on gross histopathologic examination. This can occur because of the sectioning intervals, especially if the
lesion is not easily palpable after NST. Thus there are experimental impediments to correlating DCE-MRI findings with the results of routine pathologic evaluation of breast specimens. In this study, we therefore analyzed the posttreatment parametric images in a manner similar to the analysis of histologic specimens; namely, slice by slice, every 5 mm. This approach was particularly important in cases of diffuse disease with small, scattered foci, where the exact demarcation of residual tumor tissue is more difficult. However, because of the relatively small number of cases studied, the analyses of the parametric images, as well as data concerning disease prognosis and recurrence, are not amenable to statistical evaluation.
569
CHOU ET AL
Academic Radiology, Vol 14, No 5, May 2007
Figure 5. “Three time point” parametric image of a central axial breast slice from a patient whose tumor responded to therapy (Patient 6). (a) Before neoadjuvant systemic therapy (NST). (b) After NST. Note the change in coloring (from red ⫹ green, to blue) and the presence of residual small cancerous loci (red) after NST.
Figure 6. “Three time point” parametric image of a central axial breast slice from a patient whose tumor failed to respond to therapy (Patient 15). (a) Before neoadjuvant systemic therapy (NST). (b) After NST. Note the substantial increase in tumor size and in area colored red after NST.
570
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
Table 3 Evaluation of the Segmentation and Intensity of the Regions Colored Blue in the Parametric Images Obtained Pre- and Post-NST Volume of Blue Pixels, cm3
Median Color Intensity of Blue Pixels
Case No.
Pretreatment
Posttreatment
Change (%)
Pretreatment
Posttreatment
Change (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
6.4 1.6 16.0 5.1 5.9 4.0 13.4 5.2 1.9 11.7 5.6 19.2 1.1 4.4 14.9 2.2 NA
0.3 0.3 2.4 0.9 2.8 5.0 4.2 2.2 1.2 3.4 9.7 20.6 0.4 3.4 16.0 1.3 10.5
⫺96 ⫺81 ⫺85 ⫺83 ⫺52 ⫹26 ⫺69 ⫺57 ⫺38 ⫺71 ⫹75 ⫹7 ⫺64 ⫺23 ⫹7 ⫺42 NA
134 150 146 159 102 170 147 135 155 154 134 147 162 141 159 133 NA
120 120 123 126 127 178 141 135 136 140 129 141 158 131 159 147 158
⫺10 ⫺20 ⫺16 ⫺21 ⫹25 ⫹5 ⫺4 0 ⫺12 ⫺9 ⫺4 ⫺4 ⫺2 ⫺7 0 ⫹11 NA
NST: neoadjuvant systemic therapy; NA: not available.
It was previously shown that patients with significant tumor regression after NST experienced a longer period of disease-free survival than those whose tumors were less responsive to treatment (15). As expected from the generally known prognostic factors in cases of infiltrating ductal carcinoma (35,36), all four recurrences in our study occurred in patients with Stage III cancers: three of the four patients were diagnosed with lymph node metastasis, and the fourth had developed a very large residual mass (60.9 cm3). In contrast, we found that chemotherapeutic treatment of the three patients with lymph node metastasis induced appreciable regression in tumor volume by 92%, 66%, and 30%, respectively, as well as a reduction in the color intensity of the red and green pixels, indicating an appreciable response to therapy. Thus it appears that both regression in tumor size and suppression of vascular transcapillary transfer can be used as markers for monitoring response to treatment; however, these factors may not be as effective as lymph node metastasis in predicting recurrence of disease. Notably, the fourth recurrence of carcinoma after mastectomy was of a large tumor, within which, despite NST, both the number and color intensity of red plus green pixels increased. This may suggest that quantification of the changes in the 3TP parametric images may be more effective in predicting recurrence rates
in lymph node–negative than in lymph node–positive patients. In our parametric images, the washout pattern (color hue) is related to the efflux transcapillary transfer rate constant. According to the common physiologic models applied to analyze contrast enhanced breast MRI (37), this constant is, in turn, directly proportional to the outflux transcapillary transfer constant and inversely proportional to the extravascular, extracellular volume fraction. The estimated washin rate (denoted by color intensity) is related to the influx transcapillary constant. We assumed that the influx and outflux transfer constants are equal; hence, the method was standardized in terms of the transcapillary transfer constant and extravascular, extracellular volume fraction. This and other assumptions underlying the common physiologic models (37– 40), could prove to be invalid, preventing interpretation of the color scheme in terms of the transcapillary transfer constants. Furthermore, it is clear that, because of the limited number of time points chosen, this method cannot be used to determine the exact values of the transcapillary transfer constants. Rather, it can only provide approximate assessment of these constants (Fig 1), for use in the assembly and then segmentation of similar enhancement patterns, enabling discrimination between malignant and noninvolved
571
CHOU ET AL
breast tissue. Future development of DCE-MRI, with high spatial and temporal resolution and high signal-to-noise ratio, would make it possible to construct a more comprehensive physiologic model for quantifying changes in the values of the model’s various parameters. However, only large-scale clinical studies can verify whether changes in the parameters of a more comprehensive model are, indeed, critical predictors of tumor response to therapy, and better indicators of possible disease recurrence than the approximated parameters described inhere. In summary, we have demonstrated the application of a standardized, high-resolution parametric dynamic contrast-enhanced magnetic resonance imaging method, to monitor and assess the response of breast cancer to NST. This method provides an objective means by which to segment the volume of pixels reflecting viable tumor tissue, both before and after neoadjuvant systemic therapy, and to estimate in the course of treatment changes in this volume, as well as in the vascular transcapillary transfer properties. Additional MRI data from large-scale clinical studies, for comparison with prognostic factors obtained by histopathologic and immunohistochemical assessment of tumor tissue as well as long-term follow-up of patients are necessary to validate the clinical efficacy of this method to help predict recurrence/survival rates. ACKNOWLEDGMENTS
We would like to thank Dr. Dov Grobgeld for the development of the software and Ms. Barbara Morgenstern for skillful editing. H. Degani is the incumbent of the Fred and Andrea Fallek Professorial Chair for Breast Cancer Research. REFERENCES 1. Kaufmann M, Hortobagyi GN, Goldhirsch A, et al. Recommendations from an international expert panel on the use of neoadjuvant (primary) systemic treatment of operable breast cancer: an update. J Clin Oncol 2006; 24:1940 –1949. 2. Helbich TH. Contrast-enhanced magnetic resonance imaging of the breast. Eur J Radiol 2000; 34:208 –219. 3. Padhani AR, Ah-See ML, Makris A. MRI in the detection and management of breast cancer. Expert Rev Anticancer Ther 2005; 5:239 –252. 4. Kaiser WA, Zeitler E. MR imaging of the breast: fast imaging sequences with and without Gd-DTPA. Preliminary observations. Radiology 1989; 170:681– 686. 5. Knopp MV, von Tengg-Kobligk H, Choyke PL. Functional magnetic resonance imaging in oncology for diagnosis and therapy monitoring. Mol Cancer Ther 2003; 2:419 – 426. 6. Schnall MD, Blume J, Bluemke DA, et al. Diagnostic architectural and dynamic features at breast MR imaging: multicenter study. Radiology 2006; 238:42–53. 7. Rieber A, Brambs HJ, Gabelmann A, et al. Breast MRI for monitoring response of primary breast cancer to neo-adjuvant chemotherapy. Eur Radiol 2002; 12:1711–1719.
572
Academic Radiology, Vol 14, No 5, May 2007
8. Wasser K, Sinn HP, Fink C, et al. Accuracy of tumor size measurement in breast cancer using MRI is influenced by histological regression induced by neoadjuvant chemotherapy. Eur Radiol 2003; 13:1213–1223. 9. Esserman L, Kaplan E, Partridge S, et al. MRI phenotype is associated with response to doxorubicin and cyclophosphamide neoadjuvant chemotherapy in stage III breast cancer. Ann Surg Oncol 2001; 8:549 –559. 10. Cheung YC, Chen SC, Su MY, et al. Monitoring the size and response of locally advanced breast cancers to neoadjuvant chemotherapy (weekly paclitaxel and epirubicin) with serial enhanced MRI. Breast Cancer Res Treat 2003; 78:51–58. 11. Rosen EL, Blackwell KL, Baker JA, et al. Accuracy of MRI in the detection of residual breast cancer after neoadjuvant chemotherapy. AJR Am J Roentgenol 2003; 181:1275–1282. 12. Martincich L, Montemurro F, De Rosa G, et al. Monitoring response to primary chemotherapy in breast cancer using dynamic contrast-enhanced magnetic resonance imaging. Breast Cancer Res Treat 2004; 83:67–76. 13. El Khoury C, Servois V, Thibault F, et al. MR quantification of the washout changes in breast tumors under preoperative chemotherapy: feasibility and preliminary results. AJR Am J Roentgenol 2005; 184:1499 –1504. 14. Partridge SC, Gibbs JE, Lu Y, et al. Accuracy of MR imaging for revealing residual breast cancer in patients who have undergone neoadjuvant chemotherapy. AJR Am J Roentgenol 2002; 179:1193–1199. 15. Partridge SC, Gibbs JE, Lu Y, et al. MRI measurements of breast tumor volume predict response to neoadjuvant chemotherapy and recurrence-free survival. AJR Am J Roentgenol 2005; 184:1774 –1781. 16. Delille JP, Slanetz PJ, Yeh ED, et al. Invasive ductal breast carcinoma response to neoadjuvant chemotherapy: noninvasive monitoring with functional MR imaging pilot study. Radiology 2003; 228:63– 69. 17. Chang YC, Huang CS, Liu YJ, et al. Angiogenic response of locally advanced breast cancer to neoadjuvant chemotherapy evaluated with parametric histogram from dynamic contrast-enhanced MRI. Phys Med Biol 2004; 49:3593–3602. 18. Armitage P, Behrenbruch C, Brady M, et al. Extracting and visualizing physiological parameters using dynamic contrast-enhanced magnetic resonance imaging of the breast. Med Image Anal 2005; 9:315–329. 19. Padhani AR, Hayes C, Assersohn L, et al. Prediction of clinicopathologic response of breast cancer to primary chemotherapy at contrastenhanced MR imaging: initial clinical results. Radiology 2006; 239:361– 374. 20. Pickles MD, Lowry M, Manton DJ, et al. Role of dynamic contrast enhanced MRI in monitoring early response of locally advanced breast cancer to neoadjuvant chemotherapy. Breast Cancer Res Treat 2005; 91:1–10. 21. Hayes C, Padhani AR, Leach MO. Assessing changes in tumour vascular function using dynamic contrast-enhanced magnetic resonance imaging. NMR Biomed 2002; 15:154 –163. 22. Manton DJ, Chaturvedi A, Hubbard A, et al. Neoadjuvant chemotherapy in breast cancer: early response prediction with quantitative MR imaging and spectroscopy. Br J Cancer 2006; 94:427– 435. 23. Fischer U, Zachariae O, Baum F, et al. The influence of preoperative MRI of the breasts on recurrence rate in patients with breast cancer. Eur Radiol 2004; 14:1725–1731. 24. Thibault F, Nos C, Meunier M, et al. MRI for surgical planning in patients with breast cancer who undergo preoperative chemotherapy. AJR Am J Roentgenol 2004; 183:1159 –1168. 25. Denis F, Desbiez-Bourcier AV, Chapiron C, et al. Contrast enhanced magnetic resonance imaging underestimates residual disease following neoadjuvant docetaxel based chemotherapy for breast cancer. Eur J Surg Oncol 2004; 30:1069 –1076. 26. Degani H, Gusis V, Weinstein D, et al. Mapping pathophysiological features of breast tumors by MRI at high spatial resolution. Nat Med 1997; 3:780 –782. 27. Furman-Haran E, Degani H. Parametric analysis of breast MRI. J Comput Assist Tomogr 2002; 26:376 –386. 28. Kelcz F, Furman-Haran E, Grobgeld D, et al. Clinical testing of highspatial-resolution parametric contrast-enhanced MR imaging of the breast. AJR Am J Roentgenol 2002; 179:1485–1492.
Academic Radiology, Vol 14, No 5, May 2007
BREAST CANCER RESPONSE TO CHEMOTHERAPY BY MRI
29. Furman-Haran E, Grobgeld D, Margalit R, et al. Response of MCF7 human breast cancer to tamoxifen: evaluation by the three-time-point, contrast-enhanced magnetic resonance imaging method. Clin Cancer Res 1998; 4:2299 –2304. 30. Tofts PS, Kermode AG. Measurement of the blood-brain barrier permeability and leakage space using dynamic MR imaging. 1. Fundamental concepts. Magn Reson Med 1991; 17:357–367. 31. Tofts PS. Modeling tracer kinetics in dynamic Gd-DTPA MR imaging. J Magn Reson Imaging 1997; 7:91–101. 32. Furman-Haran E, Schechtman E, Kelcz F, et al. Magnetic resonance imaging reveals functional diversity of the vasculature in benign and malignant breast lesions. Cancer 2005; 104:708 –718. 33. Therasse P, Arbuck SG, Eisenhauer EA, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000; 92:205–216. 34. Moneer M, El-Didi M, Khaled H. Breast conservative surgery: is it appropriate for locally advanced breast cancer following downstaging by neoadjuvant chemotherapy? A pathological assessment. Breast 1999; 8:315–319.
35. Singletary SE, Connolly JL. Breast cancer staging: working with the sixth edition of the AJCC Cancer Staging Manual. CA Cancer J Clin 2006; 56:37– 47; quiz 50 –31. 36. Yarbro JW, Page DL, Fielding LP, et al. American Joint Committee on Cancer Prognostic Factors Consensus Conference. Cancer 1999; 86: 2436 –2446. 37. Tofts PS, Brix G, Buckley DL, et al. Estimating kinetic parameters from dynamic contrast-enhanced T1-weighted MRI of a diffusable tracer: standardized quantities and symbols. J Magn Reson Imaging 1999; 10:223–232. 38. Yankeelov TE, Rooney WD, Li X, et al. Variation of the relaxographic “shutter-speed” for transcytolemmal water exchange affects the CR bolus-tracking curve shape. Magn Reson Med 2003; 50:1151–1169. 39. Dadiani M, Margalit R, Sela N, et al. High-resolution magnetic resonance imaging of disparities in the transcapillary transfer rates in orthotopically inoculated invasive breast tumors. Cancer Res 2004; 64:3155– 3161. 40. Hassid Y, Furman-Haran E, Margalit R, et al. Non invasive magnetic resonance imaging of transport and interstitial fluid pressure in ectopic human lung tumors. Cancer Res 2006; 66:4159.
573